[Jesse Nooijen]

Jesse:
the 4 leadership dimension are highly corelated and unavoidably so.
You may want to go PCA to sum them up and using thebresulting vector as your new dependent variable.

Hey Carlo, I am using them as IV and controls, so you would suggest I follow the PCA road? As example I would then be testing Charismatic/Value-based Leadership (IV) on revenue & current ratio (DVs) while controlling for GDP, GDP Growth, HHI, Inflation, Political Stability, and the vector of (TeamOriented, Selfprotective, Human-oriented, participative, and autonomous)?

[Carlo Lazzaro]

Jesse:
I would give PCA a shot and see if this approach eases your problem.

[Jesse Nooijen]

I got PCA working and have created new controls for leadership based on 2 components (as indicated by PCA). My RE model will now run the regression as it should, while my FE model drops Charismatic-valuebased and my 2 new components. Therefore I will proceed with RE models. Thanks for the advice Carlo!

[Jesse Nooijen]

Hey Carlo, I know this is fairly late in the process, but I was wondering on your opinion on my research design in general. I am wondering whether I made a critical design flaw since the leadership scores are on the country-level and remain static across years, hence making it more relevant to compare country-level leadership and its effects on business performance using cross-sectional analysis.

[Carlo Lazzaro]

Jesse:
as we know, due to demeaning the -fe- estimator drops all the time-invariant (dependent and independent ) variables.
That said:
1) you can go -re-;
2) you may want to consider -mundlak- is some time-invariant preditors are worth analysing.

[Jesse Nooijen]

Jesse:
as we know, due to demeaning the -fe- estimator drops all the time-invariant (dependent and independent ) variables.
That said:
1) you can go -re-;
2) you may want to consider -mundlak- is some time-invariant preditors are worth analysing.

My supervisor finally provided one line of feedback that panel data cannot have variables that do not vary over time. Given your answer, I still feel like pursuing RE as that was my initial goal. Do you still consider this an option given my static IV? I know this is similar to my last question but I just wanted to see your opinion given his remarks. I value your opinion more, as it has been very useful this psat week, which I appreciate.

[Carlo Lazzaro]

Jesse:
I assume that your supervisor meant that, if most of the variables are time-invariant, the coefficient estimates, when feasible, are poor.
That said, my previous reply still holds.
The -re- estimaytor will provide you with a coefficient for time-invariant variables too.
The only caveat rests on the evidence that, if -fe- is the way to go, -re- is inconsistent.

[Jesse Nooijen]

Jesse:
I assume that your supervisor meant that, if most of the variables are time-invariant, the coefficient estimates, when feasible, are poor.
That said, my previous reply still holds.
The -re- estimaytor will provide you with a coefficient for time-invariant variables too.
The only caveat rests on the evidence that, if -fe- is the way to go, -re- is inconsistent.

Dear Carlo, thanks again for your advice. After a lot of trial and effort I managed to get all my intended regressions to run today which I can present to my supervisor!

[Jesse Nooijen]

I am contemplating whether my current models are fit to answer my hypotheses:

●Hypothesis 1: Charismatic/Value-based leadership has a more positive influence on hotel revenue during the crisis period than during the base period.
●Hypothesis 2 Charismatic/Value-based leadership has a more positive influence on hotel current ratio during the crisis period than during the base period

●Hypothesis 3: Team-Oriented leadership has a more positive influence on hotel revenue during the base period than during the crisis period.
●Hypothesis 4: Team-Oriented leadership has a more positive influence on hotel current ratio during the base period than during the crisis period.

●Hypothesis 5: Participative leadership has a more positive influence on hotel revenue during the base period than during the crisis period.
●Hypothesis 6: Participative leadership has a more positive influence on hotel current ratio during the base period than during the crisis period

I had created the following regressions to do so, but now I doubt them;

xtreg Revenue CharismaticValuebased covid_period covid_charis HHI GDPG Inflation ln_GDP Political ln_Assets Comp1 Comp2, re robust
est store model1

xtreg Revenue TeamOriented covid_period covid_team HHI GDPG Inflation ln_GDP Political ln_Assets Comp3 Comp4, re robust
est store model2

xtreg Revenue Participative covid_period covid_part HHI GDPG Inflation ln_GDP Political ln_Assets Comp5 Comp6, re robust
est store model3

//Regressions using Current
xtreg Current CharismaticValuebased covid_period covid_charis HHI GDPG Inflation ln_GDP Political ln_Assets Comp1 Comp2, re robust
est store model4

xtreg Current TeamOriented covid_period covid_team HHI GDPG Inflation ln_GDP Political ln_Assets Comp3 Comp4, re robust
est store model5

xtreg Current Participative covid_period covid_part HHI GDPG Inflation ln_GDP Political ln_Assets Comp5 Comp6, re robust
est store model6

Where covid_charis covid_team covid_part are the interaction terms between covid dummy and my leadership variables. Given my leadership variables are static across the years, and my covid variable is a dummy, I now realize that this will not yield the most insightful results. Instead, I was contemplating whether I should create two regressions per hypothesis comparing the individual results like;

xtreg Revenue CharismaticValuebased covid_period HHI GDPG Inflation ln_Assets ln_GDP Political Comp1 Comp2 if year < 2020, re robust
est store model_1a
xtreg Revenue CharismaticValuebased covid_period HHI GDPG Inflation ln_Assets ln_GDP Political Comp1 Comp2 if year >= 2020, re robust
est store model_1b

What would your opinion be on running two separate regressions like this and comparing the coefficients and significance to draw conclusion for e.g.Hypothesis 1: Charismatic/Value-based leadership has a more positive influence on hotel revenue during the crisis period than during the base period.
I have also noticed that the explanatory power of my model is quite low with 2-3%, which I will investigate.